Authors:
N. T. Sanjaai Raajan, S. Vaishnavii, Trilok Srinivasan, S. V. Siddarth Reddy

Addresses:
Department of Computer Science and Engineering, SRM Institute of Science and Technology, Ramapuram, Chennai, Tamil Nadu, India.

Abstract:

Heavy transport mode vehicles, such as trucks and buses, use very large amounts of air braking systems to ensure operational safety. The braking mechanism - any failure can cause serious accidents, loss of life and economic damage. Conventional methods for brake maintenance are mostly reactive, based on manual inspection and scheduled servicing, and may not detect faults early. With the development of Artificial Intelligence (AI) and Machine Learning (ML), Predictive Maintenance (PdM) techniques can analyse large volumes of sensor data generated by vehicle systems to identify hidden patterns and early signs of brake deterioration. However, many AI-based systems operate as black-box models, with predictions that are hard to interpret, which can cause distrust in safety-critical environments, including transportation. To address this issue, this paper presents a real-time, explainable hybrid artificial intelligence system for detecting brake defects in heavy transport vehicles. The proposed system uses a Random Forest model for current-sample scoring and an LSTM model to analyse the latest 5 sequential samples, with the Random Forest output applied early in warmup before switching to RFLSTM late fusion. The framework further converts the fault probability into a vehicle health score and incorporates SHAP-based explanation for unbiased decision support. Multilingual Streamlit dashboard with separate Driver View and Technical View modes, recommendation messages, trend and timeline visualisation, and cooldown-controlled WhatsApp and SMS alerts, integrated to facilitate drivers and maintenance personnel during replay-based monitoring.

Keywords: Predictive Maintenance; Machine Learning; Explainable Artificial Intelligence; Brake Fault Prediction; Air Brake System; Heavy Transport Vehicles; Long Short-Term Memory.

Received on: 03/05/2025, Revised on: 28/06/2025, Accepted on: 01/10/2025, Published on: 12/06/2026

DOI: 10.69888/FTSCS.2026.000687

FMDB Transactions on Sustainable Computing Systems, 2026 Vol. 4 No. 2, Pages: 135-148

• Views : 45
• Downloads : 7

Download PDF